Prior-Auth Automation: POC to Production
How Limestone shipped supervised automation inside payer portals for a PE-backed healthcare platform
- Healthcare
- Prior Auth
- Automation
- UX/UI Design
- Industries
- Healthcare, Prior Authorization & Referral Management
- Platform
- Browser extension (Chrome, Manifest V3) + platform modernization
- Delivery scope
- Managed delivery, fixed price: automation engineering, product-wide UX audit and redesign, compliance hardening
ABOUT THE PROJECT
Project overview
- Engagement
- March 2026 — ongoing, production since June
- Team
- PM, BA, 3 full-stack developers, UX/UI designer, Design Lead
- Core technologies
- Chrome Extension (Manifest V3), React/TypeScript, Vite, client-side PDF processing
The client is a PE-backed healthcare technology company focused on prior authorization and referral management, serving hospitals, health systems, and payers. The platform carries a decade of payer-specific workflow knowledge, with a roadmap for modernization without interrupting the workflow, EHR integration, and fax operations already running.
Limestone joined in March 2026 on fixed-price managed delivery. An April proof of concept proved a browser extension could drive payer portals from inside the coordinator's own session. The team productized the extension with stakeholders in the loop at every stage, publishing to production in June.
The Challenge
Automate portals that were never meant to be automated.
-
Payer portals refuse standardization
Workflows branch on procedure codes, place of service, health plan, and clinical context. Every payer differs in authentication, data, and submission paths.
-
Automation needs a human nearby
MFA prompts, clinical questions, and unexpected portal states can't break the run. Intervention had to be part of the workflow, not a failure of it.
-
A visibility gap
High-volume workflows needed a UX overhaul to reduce friction, clarify case and payer states, and keep automation progress visible.
-
Modernize without starting over
Ten years of payer workflow knowledge and live operations had to survive the move: a V1.5 bridge toward the future platform, not a rewrite.
WHAT WE BUILT
Solution
A production automation channel, inside the coordinator's browser. The extension runs payer-specific automation within the coordinator's own authenticated session, keeps a controlled connection to the platform, and hands control to a human whenever the portal demands one.
-
Automated Execution
Case data in, predefined payer workflows executed, portal interactions navigated, outcomes captured and synchronized back to the platform.
-
Human-in-the-Loop Control
Take over and resume, predefined MFA handoffs, general handoff for unanticipated conditions, and manual result entry, all inside the same run.
-
Automation Control Center
Cross-tab orchestration with centralized visibility over every active run, so coordinators monitor and switch between concurrent executions in real time.
-
Action-Required Alerts
Native OS notifications surface intervention requests regardless of which tab or application has focus.
-
Observe Mode
Captures real activity on non-scripted portals and turns it into the foundation for new automation scripts: observe, capture, send back.
-
Execution Telemetry
Every run captured: automated completion, handoffs, and workflow-level signals, the base for measuring script performance and prioritizing the next automation.
DESIGN STREAM
The redesign that carried the automation.
An AI-driven platform redesign that unified five product areas on one token-based design system, with components and patterns reused instead of five separate looks. The payoff is consistency, faster delivery of new screens, and an interface coordinators can read at speed.
INSIDE THE COORDINATOR'S SESSION
Regulated by design, hardened before launch.
Automation executes inside the coordinator's own authenticated browser session, not a separate system. Compliance work was delivered as engineering scope: a PHI remediation pass across recorded values, errors, notifications, logs, and stored output, session and transport hardening, structured logging with PHI redaction, and HIPAA-focused security work completed ahead of the production release.
DELIVERY TIMELINE
Four beats: prove it, productise it, ship it, scale it.
-
March 2026
Engagement Start
Fixed-price managed delivery
-
April
Prove It
Live portal driven in-session
-
May
Productise It
Multi-tab · queue · integration
-
June
Ship It
Hardened · published to production
-
July+
Scale It
Observe Mode · payer library
SUPERVISED AUTOMATION
Not automation without people. Automation that knows when to ask.
-
HUMAN INPUT EXPECTED
Control passes over, and back.
MFA, clinical questions, and unexpected portal states are part of the workflow, not errors. The coordinator steps in, automation resumes, and the case never leaves the run.
-
KNOWLEDGE CARRIED FORWARD
Modernization without starting over.
The client's most valuable asset was a decade of payer workflow knowledge. Scripts were ported into the new TypeScript engine with legacy operations running throughout. Nothing was thrown away.
-
A SYSTEM, NOT A SCRIPT LIBRARY
Coverage that compounds.
Observe Mode turns real portal sessions into new automations, and telemetry on completions and handoffs shows which workflows need attention next. Each run feeds the next script.
TEN YEARS OF PAYER WORKFLOW KNOWLEDGE, CARRIED INTO A NEW ENGINE.
Impact & Results
What shipped, and what it proved.
- From proof of concept to a published production Chrome extension, establishing a new browser-based channel for payer portal automation.
- Human-in-the-loop automation became a production capability: automation handles the repeatable work while coordinators keep control for MFA, clinical questions, and manual paths.
- The Automation Control Center gives coordinators centralized visibility across concurrent runs, with native OS alerts whenever intervention is needed.
- Observe Mode made payer coverage a scalable expansion model: observe a new workflow, capture it, deploy a reusable script.
- Execution telemetry captures runs, completions, handoffs, and workflow signals, making automation measurable and improvable.
- A decade of payer workflow knowledge was extended into the new TypeScript automation engine without losing it.
- A unified design language and redesigned core experiences shipped across dashboards, work queues, case management, and automation controls.